Association for Public Policy Analysis & Management

We examine changes in cortisol patterns from a regular school week to a high-stakes testing week and whether cortisol responsivity is associated with performance on the high-stakes test. Ninety-three mostly low-income students (ages 7.9-15.6; 55% female) from three New Orleans charter schools provided six-times-daily cortisol samples in a regular school week, a low-stakes testing week, and a high-stakes testing week (N=12 samples per week). The charter network provided student grades, test scores, and demographic information. We estimate that cortisol increased 17% on average in the hour before the morning high-stakes testing period relative to the same timeframe in a regular school week. This masks variation, with an unadjusted interquartile range from -8% to +74%. Relative to those with little change in cortisol from baseline to the high-stakes week, those with a large increase or a large decrease in cortisol scored 0.30 standard deviations lower on the on the high-stakes exam. These findings are policy-relevant, given demonstrated differences in stress responsivity by socioeconomic background. If stress responses affect test performance, policymakers should carefully consider what tests really measure as they use them for high-stakes decision-making.